refactor(evaluator): split mod.rs into focused submodules

AGENTS.md

@@ -88,8 +88,11 @@ parser/
 
 // Evaluator module structure
 evaluator/
-├── mod.rs          // Main evaluation engine
-├── tests.rs        // Evaluator tests
+├── mod.rs          // Public interface: EvaluationContext, RuleMatch, re-exports
+├── tests.rs        // Unit tests for EvaluationContext and RuleMatch
+├── engine/         // Core evaluation engine submodule
+│   ├── mod.rs      // evaluate_single_rule, evaluate_rules, evaluate_rules_with_config
+│   └── tests.rs    // Engine unit tests
 ├── types.rs        // Type interpretation with endianness
 ├── strength.rs     // Strength modifier application
 ├── offset/         // Offset resolution submodule

docs/ARCHITECTURE.md

@@ -86,7 +86,8 @@ libmagic-rs/
 │   │   └── grammar.rs      # nom-based parsing combinators
 │   │
 │   ├── evaluator/          # Rule evaluation engine
-│   │   ├── mod.rs          # Main evaluation logic, EvaluationContext
+│   │   ├── mod.rs          # Public API surface with re-exports, EvaluationContext, RuleMatch
+│   │   ├── engine.rs       # Core evaluation logic (evaluate_single_rule, evaluate_rules, evaluate_rules_with_config)
 │   │   ├── offset.rs       # Offset resolution
 │   │   ├── types.rs        # Type reading with bounds checking
 │   │   ├── operators.rs    # Comparison operations

docs/src/architecture.md

@@ -126,7 +126,10 @@ The evaluator executes magic rules against file buffers to identify file types.
 
 **Structure:**
 
-- `mod.rs`: Main evaluation engine with `EvaluationContext` and `RuleMatch`
+- `mod.rs`: Public API surface (~720 lines) with `EvaluationContext`, `RuleMatch` types, and re-exports
+- `engine/`: Core evaluation engine submodule
+  - `mod.rs`: `evaluate_single_rule`, `evaluate_rules`, and `evaluate_rules_with_config` functions
+  - `tests.rs`: Engine unit tests
 - `types.rs`: Type interpretation with endianness handling and signedness coercion
 - `offset/`: Offset resolution submodule
   - `mod.rs`: Dispatcher (`resolve_offset`) and re-exports
@@ -139,6 +142,8 @@ The evaluator executes magic rules against file buffers to identify file types.
   - `comparison.rs`: `compare_values`, `apply_less_than`/`greater_than`/`less_equal`/`greater_equal`
   - `bitwise.rs`: `apply_bitwise_and`, `apply_bitwise_and_mask`, `apply_bitwise_xor`, `apply_bitwise_not`
 
+**Organization Note:** The evaluator module was refactored to split a monolithic 2,638-line `mod.rs` into focused submodules, keeping the public API surface in `mod.rs` and moving core evaluation logic to `engine/mod.rs`. This maintains the same public API through re-exports (no breaking changes) while improving code organization and staying within the 500-600 line module guideline.
+
 **Implemented Features:**
 
 - ✅ **Hierarchical Evaluation**: Parent rules must match before children

docs/src/evaluator.md

@@ -21,9 +21,22 @@ File Buffer → Offset Resolution → Type Reading → Operator Application →
 Memory Map    Context State      Endian Handling   Match Logic      Hierarchical
 ```
 
+## Module Organization
+
+The evaluator module is organized into focused submodules:
+
+- **`evaluator/engine/mod.rs`** - Core evaluation logic (`evaluate_single_rule`, `evaluate_rules`, `evaluate_rules_with_config`)
+- **`evaluator/mod.rs`** - Public API surface (types, context, re-exports)
+- **`evaluator/offset/mod.rs`** - Offset resolution
+- **`evaluator/operators/mod.rs`** - Operator application
+- **`evaluator/types.rs`** - Type reading and coercion
+- **`evaluator/strength.rs`** - Rule strength calculation
+
+From a public API perspective, all types and functions are imported from the `evaluator` module as before -- the internal organization is transparent to library users.
+
 ## Core Components
 
-### EvaluationContext (`evaluator/mod.rs`)
+### EvaluationContext
 
 Maintains state during rule processing:
 
@@ -49,7 +62,7 @@ Note: Fields are private; use accessor methods like `current_offset()`, `recursi
 - `timeout_ms()` - Query configured timeout
 - `reset()` - Reset context state for reuse
 
-### RuleMatch (`evaluator/mod.rs`)
+### RuleMatch
 
 Represents a successful rule match: